Sarah Cliston, David M. Berry, Anthony C. Hay, Mark C. Marino, Peter Millican, Arthur I. Schwarz, Jeff Shrager, and Peggy Weil. Inventing Eliza. The MIT Press. 2026. ISBN13: 978-0-262-05248-1.
Society has lost many valuable pieces of computer history. Digital media fails, websites go offline, and personal memories fade. Remembering or reconstructing the early stages of digital computing in the 1950s and 1960s is particularly difficult.
In Inventing ELIZA from The MIT Press, a team of eight authors created a valuable multi-lens analysis of the now-famous program ELIZA. The program, which was technically the ELIZA script running within the DOCTOR framework, simulated human nature and maintained a conversation by adopting the approach of a Rogerian psychotherapist. The source code of later versions of the program written in BASIC and other languages were available, but the code behind the original 1965 version of ELIZA had been lost.
ArnoldReinhold, CC BY-SA 3.0 <https://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons
Thanks to good luck during an archival search, a printout of the original MAD-SLIP code was found, allowing the author team to dig into the code and its implications. After an insightful introduction by Janet H. Murray, author of Hamlet on the Holodeck, the team starts their analysis with a biography of ELIZA’s developer Joseph Weizenbaum. Weizenbaum escaped from Germany in the 1930s, eventually ending up at MIT in 1963. His previous work on the Symmetric List Processor (SLIP) programming language offered the ability to manage text in ways not easily done before. Weizenbaum used those capabilities to create ELIZA.
After the biography, the authors devote a chapter to a close reading of the ELIZA code. And by “close reading” I mean a block-by-block, and in some cases line-by-line, interpretation of the code. I’m a scrappy coder at best and have never worked in the FORTRAN-based SLIP, which was overtaken by LISP for this type of work, but SLIP offered some distinct advantages contemporary programmers take as given. First, the MAD-SLIP language could handle text as symbols. Second, the language could dynamically allocate resources at runtime. MAD-SLIP also had double-linked lists, which meant that not only was there a memory pointer to the next item in the list, there was also a pointer to the previous item. Perhaps most importantly, dynamic allocation meant lists could contain other lists—most languages did not permit this type of recursion.
Throw in ridiculously small memory and storage capacity for programs entered using punch cards and it’s a miracle anything of substance got done at all.
ArnoldReinhold, CC BY-SA 3.0 <https://creativecommons.org/licenses/by-sa/3.0>, via Wikimedia Commons
After the technical analysis, the authors turn to the implications of ELIZA on human-computer interaction at the time of its release, how the mores and attitudes of the 1960s and 1970s affected the program’s construction and interpretation, and how ELIZA and similar programs such as MrMind gave rise to the chatbots and Large Language Models (LLMs) in use today. LLMs such as ChatGPT provoke concern because their probabilistic text generation will inevitably lead to inaccuracies. LLMs don’t have built-in calculation engines so their performance on math problems is even more error-prone.
At a deeper level, the so-called ELIZA Effect, where humans anthropomorphized the computer and sought, in the broader sense, to offload decisions to the perceived impartiality of an algorithm. As Weizembaum wrote:
“Science promised [humans] power. But, as so often happens when people are seduced by promises of power, the price exacted in advance and all along the path, and the price actually paid, is servitude and impotence.” [Weizembaum, Computer Power and Human Reason, p. 227]
NW7US, CC BY-SA 4.0 <https://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons
It’s humbling to realize the inventors of human-interactive text generation systems identified the power and danger inherent in those tools so soon after their creation.
Inventing ELIZA is a groundbreaking book that serves the fields of computing history, sociology, gender studies, and computer-human interaction in equal measure. I’m glad I read it and enthusiastically recommend it.